Skip to content

Build AI methods with your coding agent, then run them anywhere: as an MCP for chatbots, as a webapp for people, or via API for your software.

kill PipeCompose

thomashebrard5d180fbe67d0
diff --git a/pipelex/libraries/pipelines/builder/builder.plx b/pipelex/libraries/pipelines/builder/builder.plx
index 26f99ecf6..19203113a 100644
--- a/pipelex/libraries/pipelines/builder/builder.plx
+++ b/pipelex/libraries/pipelines/builder/builder.plx
@@ -83,7 +83,6 @@ Available pipe operators:
That means that in the MAIN pipeline, the prompt MUST NOT be considered as an input. It should be the output of a step that generates the prompt.
- PipeOcr: A pipe that uses an OCR technology to extract text from an image or a pdf.
VERY IMPORTANT: THE INPUT OF THE PIPEOCR MUST BE either an image or a pdf or a concept which refines one of them.
-- PipeCompose: A pipe that uses Jinja2 to render a template.
Be smart about splitting the workflow into steps (sequence or parallel):
@@ -205,7 +204,6 @@ Available pipe operators:
- PipeImgGen: A pipe that uses an AI model to generate an image.
VERY IMPORTANT: IF YOU DECIDE TO CREATE A PipeImgGen, YOU ALSO HAVE TO CREATE A PIPELLM THAT WILL WRITE THE PROMPT, AND THAT NEEDS TO PRECEED THE PIPEIMGEN, based on the necessary elements.
That means that in the MAIN pipeline, the prompt MUST NOT be considered as an input. It should be the output of a step that generates the prompt.
-- PipeCompose: A pipe that renders a Jinja2 template.
- PipeOcr: A pipe that extracts text from an image or a pdf. PipeOcr must have a exactly one input which must be either an `Image` or a `PDF`.
Be smart about splitting the workflow into steps (sequence or parallel):
diff --git a/pipelex/libraries/pipelines/builder/pipe/pipe_design.plx b/pipelex/libraries/pipelines/builder/pipe/pipe_design.plx
index 264b8a08c..39f1a1c0d 100644
--- a/pipelex/libraries/pipelines/builder/pipe/pipe_design.plx
+++ b/pipelex/libraries/pipelines/builder/pipe/pipe_design.plx
@@ -12,7 +12,7 @@ PipeSequenceSpec = "A structured spec for a pipe sequence."
# Pipe operators
PipeFuncSpec = "A structured spec for a pipe func."
PipeImgGenSpec = "A structured spec for a pipe img gen."
-PipeComposeSpec = "A structured spec for a pipe jinja2."
+# PipeComposeSpec = "A structured spec for a pipe jinja2."
PipeLLMSpec = "A structured spec for a pipe llm."
PipeOcrSpec = "A structured spec for a pipe ocr."
PipeFailure = "Details of a single pipe failure during dry run."
@@ -33,7 +33,7 @@ PipeCondition = "detail_pipe_condition"
PipeLLM = "detail_pipe_llm"
PipeOcr = "detail_pipe_ocr"
PipeImgGen = "detail_pipe_img_gen"
-PipeCompose = "detail_pipe_compose"
+# PipeCompose = "detail_pipe_compose"
# ────────────────────────────────────────────────────────────────────────────────
# PIPE CONTROLLERS
@@ -160,6 +160,61 @@ This PipeCompose is part of a larger pipeline:
You will specifically generate the PipeCompose related to this signature:
@pipe_signature
+
+You can ONLY USE THE INPUTS IN THIS PIPE SIGNATURE.
+
+Here are the Jinja2 filters that are supported:
+abs — Returns the absolute value of a number.
+attr — Gets an attribute from an object, returning undefined if missing.
+batch — Groups items into sublists of a given size, optionally filling blanks.
+capitalize — Uppercases the first character and lowercases the rest.
+center — Centers a string within a given width.
+default — Returns a fallback value if the input is undefined (or falsey if enabled).
+dictsort — Sorts a dict and yields (key, value) pairs by key or value.
+escape — HTML-escapes &, <, >, ' and ".
+filesizeformat — Formats bytes as human-readable sizes (kB/MB or KiB/MiB).
+first — Returns the first item of a sequence.
+float — Converts a value to float with an optional default.
+forceescape — Forces HTML escaping even if marked safe.
+format — Applies printf-style string formatting.
+groupby — Groups items by an attribute (sorted first).
+indent — Indents each line of a string by a given width.
+int — Converts a value to int with base and default support.
+items — Iterates over a mapping’s items, empty if undefined.
+join — Concatenates sequence items with an optional separator or attribute.
+last — Returns the last item of a sequence.
+length — Returns the number of items (alias: count).
+list — Converts the value to a list (strings become lists of characters).
+lower — Converts a string to lowercase.
+map — Extracts an attribute or applies a filter across a sequence.
+max — Returns the largest item, optionally by attribute.
+min — Returns the smallest item, optionally by attribute.
+pprint — Pretty-prints a value for debugging.
+random — Returns a random item from a sequence.
+reject — Filters out items where a test passes.
+rejectattr — Filters out items based on an attribute test.
+replace — Replaces occurrences of a substring with another.
+reverse — Reverses a string or iterable.
+round — Rounds a number with precision and method (common/ceil/floor).
+safe — Marks a string as safe and prevents auto-escaping.
+select — Keeps only items where a test passes.
+selectattr — Keeps only items whose attribute passes a test.
+slice — Slices items into N columns (lists of lists), with optional fill.
+sort — Sorts an iterable, with reverse, case, and attribute options.
+string — Converts to string while preserving Markup safety.
+striptags — Removes HTML/XML tags and collapses whitespace.
+sum — Sums numbers or an attribute across items, with a start value.
+title — Title-cases the string (each word capitalized).
+tojson — Serializes a value to JSON and marks it safe for HTML.
+trim — Strips leading and trailing characters (default: whitespace).
+truncate — Shortens text to a length with optional word-safe ellipsis.
+unique — Yields unique items in first-seen order, optional attribute.
+upper — Converts a string to uppercase.
+urlencode — URL-encodes a string or builds a query string from pairs.
+urlize — Converts URLs/emails in text into clickable links.
+wordcount — Counts words in a string.
+wordwrap — Wraps text to a given width with options for breaks.
+xmlattr — Builds safe HTML/XML attributes from a dict.
"""
# ────────────────────────────────────────────────────────────────────────────────